I installed Corral and have been playing around with VMs.
I've observed an interesting problem when running some simple network tests and I'm wondering if anyone has seen similar results in their testing.
All tests are running on a 10G Network and involve running the following 2 commands.
Test 2:
Next, start a single VM (any VM) with a bridged network mode device (same network - no vlan) on the FreeNAS host and run the same test (no direct interaction with the VM)
Test 3:
Next, shutdown the VM with the bridged network mode device (same network - no vlan) on the FreeNAS host and run the same test.
Test 4:
Next, start a VM with a NAT network mode device on the FreeNAS host and run the same test.
Current conclusion? Bridged VMs severely impact network performance on the host?
Some further notes:
I called out no-vlan above because initial testing is suggesting I can't get back to the full wire-speed once I've started a VM with a bridged vlan enabled network. Shutting down the VM will still result in poor iperf performance numbers until i reboot the Host machine.
I'm also seeing poor VM to VM network performance using bridged vlan enabled network devices, which might not be surprising given the above results.
Cheers.
I've observed an interesting problem when running some simple network tests and I'm wondering if anyone has seen similar results in their testing.
All tests are running on a 10G Network and involve running the following 2 commands.
- On FreeNAS Corral Host: iperf -s
- On a remote OS targeting above host: iperf -c XXX.XXX.XXX.XXX -P1 -i1 -t100
- Result: 9.41 Gbits/sec
Test 2:
Next, start a single VM (any VM) with a bridged network mode device (same network - no vlan) on the FreeNAS host and run the same test (no direct interaction with the VM)
- Result: ~3.2 Gbits/sec
Test 3:
Next, shutdown the VM with the bridged network mode device (same network - no vlan) on the FreeNAS host and run the same test.
- Result: 9.41 Gbits/sec
Test 4:
Next, start a VM with a NAT network mode device on the FreeNAS host and run the same test.
- Result: 9.41 Gbits/sec
Current conclusion? Bridged VMs severely impact network performance on the host?
Some further notes:
I called out no-vlan above because initial testing is suggesting I can't get back to the full wire-speed once I've started a VM with a bridged vlan enabled network. Shutting down the VM will still result in poor iperf performance numbers until i reboot the Host machine.
I'm also seeing poor VM to VM network performance using bridged vlan enabled network devices, which might not be surprising given the above results.
Cheers.